Prognostic Analysis Of Lung Cancer Patients With ALK Mutation Based On Radiomics

Non-small cell lung cancer is a common type of lung cancer and has a very high lethal rate worldwide.The use of the targeted drug,Crizotinib,greatly prolongs the survival of patients with disease progression,but patients often develop resistance after the drug adaptation period.How to make full use of individual differences,according to the specific circumstances of distinct individuals for risk prediction,gathering the dominant population to carry out treatment is of great significance.At the same time,with the rapid development of imaging technology and the ever-increasing amount of data,radiomics,which can perform underlying analysis of the information provided by image data with the aid of universal imaging technology,has emerged.Based on the above background,this dissertation analyzes the progression-free survival of non-small cell lung cancer patients with positive ALK mutation.First of all,the study of image radiomic features mainly consists of four types of features,including the first-order statistical features describing the image intensity within the region of interest,and the second type,shape features that characterize the shape and size of the lesion region in three-dimensional space.The third type,texture features based on gray level co-occurrence matrix and grey level run-length method describe relative position information between different gray levels.And the last type of features decoupled by decomposition are wavelet features.From the collected 85 cases of clinical data,485 radiomic features were extracted from each case.Next,dimension reduction and modeling of the extracted high-dimensional features is performed.The Cox proportional hazards model and the lasso-cox model with characteristic dimensionality reduction are studied.Because the features are often accompanied by redundancy,the features selected by the lasso model are not completely consistent with the actual core features.The model is improved,a pre-excluded improvement model is proposed,and the Mann-Whitney U test is used to exclude unrelated features.Then using lasso-cox model to complete dimensionality reduction and modeling process.The improvement results is qualitatively and quantitatively evaluated,and proved that the features selected after the improvement are more stable and the prediction accuracy is higher.Finally,the progression-free survival analysis was performed based on the previously constructed radiomic signature model.The patients were divided into high and low risk groups.Kaplan-Meier method was used to estimate the survival rate and verify the label.In order to eliminate the confounding effects of clinical risk factors,a stratified analysis was performed to verify the validity of the radiomic signature.Then the univariate analysis of clinical risk factors and imaging histology labels are launched,after using appropriate hypothesis testing methods,the meaningful factors from univariate analysis are adopted for multivariate analysis to construct regression method.Afterwards,combine model are constructed by radiomic signature and clinical risk factors.The mixed model and the clinical model consisting only of meaningful clinical risk factors are compared and found that the consistency coefficient of the mixed model was 0.7173 and 0.6671 in the training set and testing set respectively,whose prediction accuracy is the highest.An easy to use nomogram is drawn based on the mixed model and complete the verification.